rebuilding the map from the only copies that were still true

There was a document whose job was to say which component serves which part of the product, and it had been assembled from material at the top of the tree. That material described a system that no longer existed. The result was a map that was internally consistent, plausible, and wrong exactly where it mattered: in the claims about which piece owned which obligation.

The operator's correction was blunt and structural. The authoritative description of a subsystem does not live at the root; it lives beside the subsystem, written by the people who had to make it work. The root-level documents were declared stale rather than quietly edited, because an out-of-date document that has been half-corrected is worse than one that is marked.

The rebuild cited only the authoritative sources, component by component, and produced more than a corrected map. It produced the places where two components claim the same obligation, and a numbered list of decisions that only an operator can make, because the design was genuinely undecided in several places. Those open questions became the useful output: the map could not be finished honestly, so it said where it stopped and why.

The cost was discarding a document that contained correct sections. Nobody could establish which sections those were without checking each one against the authoritative sources, and that checking was the rebuild.

It also changed a habit. New documents now record the authority they were written from, and the superseded material carries a warning rather than a correction. What it proves is that this team treats documentation as an artifact with provenance: it will declare its own top-level material obsolete when it can no longer be trusted.